Some AI experimentation

[glorybound]

Yes, that is the key. Subtle improvements. Trees and bushes usually work well, but I haven't had luck with grass. What prompt did you use?

[Pixelcon]

Well I'm sold. Beats photobashing or going full mental in 3D for detailing.

Very nice. The bushes look great too!

[Vizioen]

I would say thanks but it feels like cheating in a way.

[glorybound]

When it becomes a paid service, most hobbyists and fly-by-nighters will fall away. It'll become a tool that the rest of us use. Hopefully, it is something that can't be cracked.

[fixeighted]

I would say thanks but it feels like cheating in a way.

I would argue that if that is cheating, then so is the use of anything that isn't scratch built/textured by the individual.

Although I have the skills to produce pretty much anything alone, the time saved by now not necessarily having to is valuable, so personally I have no problem
using such tools, as they are merely a part of the toolbox for me.
Like having a pre-made scene I can buy, tweak to my liking quickly, then drop my bespoke model into.
That process alone keeps me profitable/employable

Artists many years ago were seen as having cheated by some, when they used a camera obscura...

[AEBI_VINCENT]

Hi there,

Just wanted to show some experimentation with AI. Interesting stuff but I can't imagine I'll be overhauling my workflow just yet. It certainly is going to be handy for replacing certain elements like vegetation and 3D people.
Using stable-diffusion and many many many different settings for each part of the image.
Took a solid day to get it from raw-render to the "finished" image (which was quite a bit longer than it took me to do the traditionally-created image in PS with lots of photobashing and matte-painting. Surely this can be sped up with more experience and I'm not at all satisfied with the result vs my original version.
I imagine when these AI interfaces can make better use of the information in the 3D scene it'll be much more powerful. It's a bit hard to wrangle the correct scale of elements it generates.

*Shameless plug* I'll be going over this in-depth on my Patreon next month.

May I ask how you tell Stable Diffusion how to use the "same" mountains, but just enhance their quality? I'm using your raw render for testing, but it generates completely different mountains every time. Is there a special prompt for this?

[Vizioen]

May I ask how you tell Stable Diffusion how to use the "same" mountains, but just enhance their quality? I'm using your raw render for testing, but it generates completely different mountains every time. Is there a special prompt for this?

Did you read the thread that follows?

[Pixelcon]

May I ask how you tell Stable Diffusion how to use the "same" mountains, but just enhance their quality? I'm using your raw render for testing, but it generates completely different mountains every time. Is there a special prompt for this?

Well you could sign up for the free trial and watch me run thru what I did here: https://www.patreon.com/posts/april-...al-2-101030003

But in essence make sure you're using img2img, and you're using a denoise between .35-.6 and use controlnets to make sure elements of the underlying image is retained. Namely canny and linework. Sometimes normal and depth can help too. Play with their weights too.
I would explain more, but it would take hours :P

[AEBI_VINCENT]

Did you read the thread that follows?

Thank your for asking. Yes, I followed the entire thread, which led me to install SD and Controlnets. However, my specific question was not answered until now.

Well you could sign up for the free trial and watch me run thru what I did here: https://www.patreon.com/posts/april-...al-2-101030003

But in essence make sure you're using img2img, and you're using a denoise between .35-.6 and use controlnets to make sure elements of the underlying image is retained. Namely canny and linework. Sometimes normal and depth can help too. Play with their weights too.
I would explain more, but it would take hours :P

I am unfamiliar with most of these settings, looks like I require some more knowledge and testing. I will consider watching your tutorial and I appreciate your help.

[kosso_olli]

Well I'm sold. Beats photobashing or going full mental in 3D for detailing.

Is this one done with Magnific?

[Vizioen]

Yes it's a combination of a couple of different outputs though so quite some masking and brushing involved but speeds up everything by a lot.

[glorybound]

Well you could sign up for the free trial and watch me run thru what I did here: https://www.patreon.com/posts/april-...al-2-101030003

But in essence make sure you're using img2img, and you're using a denoise between .35-.6 and use controlnets to make sure elements of the underlying image is retained. Namely canny and linework. Sometimes normal and depth can help too. Play with their weights too.
I would explain more, but it would take hours :P

I there a part one? I see that Google Mas is part one.

[glorybound]

ControlNet is just a plugin to get a GUI of SD in PS and nothing more. Using PS in the browser is the same thing.

[Vizioen]

ControlNet is just a plugin to get a GUI of SD in PS and nothing more. Using PS in the browser is the same thing.

No it's not: ControlNet: A Complete Guide - Stable Diffusion Art (stable-diffusion-art.com)

[glorybound]

Yes, it adds extra conditions inside of Stable Diffusion. However, I was told that I wasn't using SD if I wasn't using ControlNet.